MAINT - change the patch_display function so that it uses the minimal TableReport

[rcap107]

skrub allows to replace the default repr of pandas and polars dataframes with the TableReport by using skrub.patch_display. This functionality can be useful, but when a dataframe is big the generation of the report takes a long time, which can cause friction when developing.

The data ops have a similar issue, and there is mitigated by using a minimal version of the table report that does not compute the associations and that does not generate the distribution plots. I think that patch_display should use this version of the table report by default, and if users want to see the associations/plots they can just call TableReport(df) manually and get the result they need.

[fatiben2002]

Hello, I will work on this one !

[rcap107]

Hi @fatiben2002, thanks for picking this up. The change itself is pretty minor, but it's hidden in a fairly deep part of the code.

The patching code is in skrub/_reporting/_patching.py, and the patching function to modify is _patch().
Then, to set the resulting TableReport to minimal mode you will need to call the method _set_minimal_mode on the TableReport. The current code is using a lambda function, and that should be modified so that it looks more like this

    def _patched(df):

        ...

        report = TableReport(
            df,
            verbose=verbose,
            max_plot_columns=max_plot_columns,
            max_association_columns=max_association_columns,
        )
        report._set_minimal_mode()
        return getattr(report, method_name)()

    setattr(cls, method_name, _patched)

(note that it's missing part of the actual code so it can't be copypasted)

After changing that, you should make sure that the tests are passing properly with the new minimal mode: this means testing that the association and distribution tabs are not present anymore in the html of the resulting TableReport.

A note of caution about this: #1907 is also touching some of the same lines of code, so there may be conflicts that pop up as you work on this and the other PR gets merged. You can look at the code that is being modified in #1907 to have an idea of how it changes the _patch function. Of course, if there are conflicts that look to complex to solve I can help with dealing with that.

This is an issue where the fix is easy to implement, but the context of the fix is advanced, so it deserves some additional care.